GPU-Accelerated Parallel FDTD on Distributed Heterogeneous Platform

Table 4

Speedup ratio (1 NP, problem size 512 × 256 × 256).

Device	Loop time	Speedup
Pure CPU	1.3541	1
Pure GPU	0.1028	13.2
CPU + GPU	0.0903 (cuRatio = 0.84)	15.0